Open 1556900941lizerui opened 2 hours ago
Describe your problem
When creating a new knowledge base, will a new idxnm be used? I found in search.py that the program will use filter to exclude the specified kb_ids. By consulting the materials, I found that when calculating similarity scores using keywords, the entire index's docFreq and termFreq are still used for score calculation. Is this problematic?
I'm sorry, I'm using the version released at the time. I checked the latest code and found that now it will combine idx and kb_id into a new table and use a table-level replacement for the original scheme. Is this different from the original approach, and will the situation I mentioned above still occur?
Describe your problem
When creating a new knowledge base, will a new idxnm be used? I found in search.py that the program will use filter to exclude the specified kb_ids. By consulting the materials, I found that when calculating similarity scores using keywords, the entire index's docFreq and termFreq are still used for score calculation. Is this problematic?